金融经济学专题：数据科学与时间序列模型的应用---以金融股票预测与经济数据运动行为研究为例

阅读 861

2023-11-20

　　开始日期: 2023-12-16

　　课时安排: 7周在线小组科研学习+5周不限时论文指导学习

　　适合人群

　　适合年级 (Grade): 高中生/大学生

　　适合专业 (Major): 应用数学、金融经济学、宏观经济学、计量经济学、金融数据分析、股票投资、商业分析等专业或希望修读相关专业的学生;学生需具备随机变量、概率论等相关知识并熟练掌握R语言。

　　导师介绍

　　Peter

　　麻省理工学院 (MIT)终身教职

Peter

　　Peter 导师以优异的成绩获得哈佛大学(Harvard University)应用数学学士学位，并当选为Phi Beta Kappa Alpha Chapter的成员。后续他攻读统计学，获得了帝国理工学院(Imperial College London)的硕士学位以及加州大学伯克利分校(University of California Berkeley)的博士学位。Peter 曾任哈佛大学统计系教授，任教期间获得了美国国家科学基金会的博士后数学科学研究奖学金。随后成为麻省理工学院Sloan管理学院终身教授兼首席研究科学家，在经济和管理科学计算研究中心(CCREMS)和国际金融服务研究中心(IFSRC)进行研究。他是风险管理项目组的积极成员，并开发了纳入行业标准RiskMetrics方法论的分析方法。现为麻省理工学院数学系金融数学与统计讲师。2014年在北京交通大学全球暑期学校任教期间，被聘为计算机与信息技术学院特聘教授。

　　自1992年以来，Peter 一直通过他的公司Kempthorne Analytics, Inc. 为各种机构提供金融和统计分析咨询服务。客户包括花旗银行(Citibank)、Colonial/Liberty Funds、美国运通(American Express)、巴黎国家银行(Banque Nationale de Paris)、佳能(Canon)等。主要项目包括：股票市场的资产选择建模、风险管理的统计分析、风险管理软件的设计和实现、衍生品定价的金融分析、灾难性风险分析--风险暴露建模和保险定价方案、股票市场以数据微观结构建模以及交易系统的设计、开发、实现。自1995年以来，Peter一直担任投资经理，利用先进的统计分析来管理各种投资项目，他在一家完全系统化的量化对冲基金IKOS CIF, LTD，担任投资组合经理和高级研究员。管理和增强投资组合的构建，alpha模型的评估、开发、执行分析和投资组合优化，以及期货和货币投资组合的风险建模和管理。作为高级研究员，他担任研究指导委员会主，管理和指导研究人员，并协调IKOS/牛津大学博士实习生计划。他于1995年联合创立了Chronos Asset Management，于1996年联合创立了Summa Capital Management。作为这两家投资管理公司的负责人，他运用自己专有的分析方法开发统计交易模型和交易系统，并监督交易操作。Peter导师持有Series 3和Series 65许可证，并在the National Futures Association是注册商品交易顾问。他活跃于John Bertram House Inc.(1998-2010)和Lynn Home for Young Women, Inc.(2005-2010)的董事会，曾担任两家非营利公司的财务主管，并担任监督信托资产管理的财务委员会主席。

　　Professor Peter received his B.S. in Applied Mathematics from Harvard University with honors and was elected a member of the Phi Beta Kappa Alpha Chapter. He pursued graduate studies in Statistics receiving M.Sc degree from Imperial College London and Ph.D from University of California Berkeley. Peter was awarded a National Science Foundation Postdoctoral Research Fellowship in Mathematical Sciences while a professor in the Department of Statistics at Harvard University. He then became an Associate Professor and Principal Research Scientist at MIT Sloan School of Management, conducting research at the Center for Computational Research in Economic and Management Sciences (CCREMS) and the International Financial Services Research Center (IFSRC). He is a member of the Risk Management Project team and has developed analytical methods of Risk Metrics methodology which is incorporated into industry standard. He is currently a Lecturer in MIT Mathematics Department on Financial Mathematics and Statistics. In 2014, while teaching at the Global Summer School of Beijing Jiaotong University, he was appointed as a special professor of the School of Computer and Information Technology.

　　Peter has been providing financial and statistical analysis consulting services to various institutions through his own company Kempthorne Analytics, Inc. since 1992, clients include Citibank, Colonial/Liberty Funds, American Express, Banque Nationale de Paris, Canon. The main projects include: asset selection modeling of stock market, statistical analysis of risk management, design and implement risk management software, financial analysis of derivatives pricing, catastrophic risk analysis--risk exposure modelling and insurance pricing schemes, stock market based on data microstructure modeling and trading system design, development and implementation. Peter has been an investment manager since 1995, using advanced statistical analysis to manage various investment projects, where he is a portfolio manager and senior Research fellow at IKOS CIF, LTD. Manage and enhance portfolio construction, evaluation, development, performance analysis and portfolio optimization of alpha models, and risk modeling and management of futures and currency portfolios. As a Senior Research Fellow, he serves on the Research Steering Committee, manages and directs researchers, and coordinates the IKOS/ Oxford PhD Internship programme. He co-founded Chronos Asset Management in 1995 and Summa Capital Management in 1996. As head of these two investment management firms, he applied his proprietary analytical methods to develop statistical trading models and trading systems, and oversaw trading operations. Peter holds Series 3 and Series 65 licenses and is a registered Commodity Trading Advisor with the National Futures Association. He was active on the boards of John Bertram House Inc.(1998-2010) and Lynn Home for Young Women, Inc.(2005-2010) and served as Treasurer of two non-profit corporations, also chaired the Finance Committee, which oversaw the management of trust assets.

　　任职学校

　　麻省理工学院(MIT)创立于1861年，是世界著名私立研究型大学，在2023年U.S.News世界大学排名中综排位列第二。学校孕育了97位诺贝尔奖得主、59位美国国家科学奖章获得者，以及75位麦克阿瑟奖获得者。

　　项目背景

　　时间序列是指将某种现象某一个统计指标在不同时间上的各个数值，按时间先后顺序排列而形成的序列。时间序列法是一种定量预测方法，亦称简单外延方法，在统计学中作为一种常用的预测手段被广泛应用。时间序列分析在第二次世界大战前应用于经济预测。二次大战中和战后，在军事科学、空间科学、气象预报和工业自动化等部门的应用更加广泛。时间序列分析(Time series analysis)是一种动态数据处理的统计方法。该方法基于随机过程理论和数理统计学方法，研究随机数据序列所遵从的统计规律，以用于解决实际问题。时间序列构成要素是：现象所属的时间，反映现象发展水平的指标数值。

　　项目介绍

　　本课程将重点介绍时间序列分析的基本方法和模型及其在经济、金融数据分析中的应用。本课程将融合计算机编程的R语言辅助时间序列模型在金融经济数据中的处理分析。目前，主流经济数据分析往往会以图形方法来进行呈现，这些可视化方法被用于大数据探索、分析模型的有效性验证和数据预测结果的展现。在本课程中，导师开发并应用了趋势和季节性的重要时间序列模型，包括经典分解和多级指数平滑模型。同时导师将利用真实世界的时间序列数据(包括美国联邦储备局、世界银行和雅虎金融数据库)对本课程中涵盖的统计概率方法进行分析和实践应用。Introduction to fundamental methods and models of time series analysis with applications in economics, finance, and public health. The course uses R to forecast time series. Graphical methods are emphasized for data exploration, analyzing the validity of models, and presenting forecast results. Important models of trend and seasonality are developed and applied, including classical decompositions and multi-stage exponential smoothing. Real-world time series data are collected from the internet and analyzed with the methods covered in the course.

　　项目大纲

　　时间序列分析导论 Introduction to Time Series Analysis

　　时间序列模型;金融时间序列 Simple Time Series Models; financial time series

　　预估噪声序列的时间序列相关性检验固定的流程 Testing estimated noise sequences for time series dependence; stationary processes

　　回归(AR)、移动平均(MA)和ARMA模型 ;模型选择和预测 Auto-regression (AR), moving average (MA), and ARMA models;model selection and forecasting

　　学术研讨1 Final Project Phase I

　　学术研讨1 Final Project Phase II

　　项目回顾和成果展示 Program Review and Presentation

　　论文辅导Project Deliverables Tutoring

　　项目收获

　　7周在线小组科研学习+5周不限时论文指导学习共125课时

　　项目报告